From local to non-local dependencies Unbounded Dependency - PowerPoint PPT Presentation

From local to non-local dependencies Unbounded Dependency Constructions (UDCs) in HPSG • A head generally realizes its arguments locally within its head domain. • Certain kind of constructions resist this generalization, such as, for example, the wh -questions discussed below. Introduction to HPSG • How can the non-local relation between a head and such arguments be 26. Mai 2009 licensed? How can the properties be captured? Kordula De Kuthy 1 2 A first example: Wh -elements Different categories can be extracted: (3) a. Which man did you talk to ? NP Wh -elements can have different functions: b. [ To [ which man ]] did you talk ? PP c. [ How ill ] has the man been ? AdjP (1) a. Who did Hobbs see ? Object of verb d. [ How frequently ] did you see the man ? AdvP b. Who do you think saw the man? Subject of verb c. Who did Hobbs give the book to ? Object of prep This sometimes provides multiple options for a constituent: d. Who did Hobbs consider to be a fool? Object of obj-control verb (4) a. Who does he rely [ on ] ? Wh -elements can also occur in subordinate clauses: b. [ On whom ] does he rely ? Unboundedness: (2) a. I asked who the man saw . b. I asked who the man considered to be a fool . (5) a. Who do you think Hobbs saw ? c. I asked who Hobbs gave the book to . b. Who do you think Hobbs said he saw ? d. I asked who you thought saw Hobbs. c. Who do you think Hobbs said he imagined that he saw ? 3 4

Unbounded dependency constructions Strong UDCs An overt constituent occurs in a non-argument position: Topicalization: An unbounded dependency construction (6) Kim i , Sandy loves i . – involves constituents with different functions – involves constituents of different categories Wh -questions: – is in principle unbounded (7) I wonder [ who i Sandy loves i ] . Wh -relative clauses: (8) This is the politician [ who i Sandy loves i ] . Two kind of unbounded dependency constructions (UDCs) – Strong UDCs It -clefts: (9) It is Kim i [ who i Sandy loves i ] . – Weak UDCs Pseudoclefts: (10) [ What i Sandy loves i ] is Kim i . 5 6 Weak UDCs Some properties of UDC constructions No overt constituent in a non-argument position: Link between filler and gap with category information needed: Purpose infinitive ( for - to clauses): (15) a. Kim i , Sandy trusts i . (11) I bought it i for Sandy to eat i . b. [ On Kim ] i , Sandy depends i . Tough movement: (16) a. * [ On Kim ] i , Sandy trusts i . (12) Sandy i is hard to love i . b. * Kim i , Sandy depends i . Relative clause without overt relative pronoun: (13) This is [ the politician ] i [ Sandy loves i ] . It -clefts without overt relative pronoun: (14) It is Kim i [ Sandy loves i ] . 7 8

Using the feature slash And this link has to be established for an unbounded length: (17) a. Kim i , Chris knows Sandy trusts i . b. [ On Kim ] i , Chris knows Sandy depends i . To account for UDCs, we will use the feature slash (so-named because it (18) a. * [ On Kim ] i , Chris knows Sandy trusts i . comes from notation like S/NP to mean an S missing an NP) b. * Kim i , Chris knows Sandy depends i . • This is a non-local feature which originates with a trace, (19) a. Kim i , Dana believes Chris knows Sandy trusts i . b. [ On Kim ] i , Dana believes Chris knows Sandy depends i . • gets passed up the tree, (20) a. * [ On Kim ] i , Dana believes Chris knows Sandy trusts i . b. * Kim i , Dana believes Chris knows Sandy depends i . • and is finally bound by a filler 9 10 An example for a strong UDC S f h The bottom of a UDC: Traces NP S c h John i   word NP VP �� phon     h c  local    1 we   � �   V � � S inherited | slash synsem  1    nonloc     � loc | cat | subcat �� � to-bind | slash {} know c h 1 NP VP • phonologically null, but structure-shares local and slash values � � � � loc | cat | subcat 1 • we’ll talk about to-bind later she h c 2 NP V � � � � loc | cat | subcat 1 , 2 12 likes i

Traces The middle of a UDC: The Nonlocal Feature Principle (NFP) Because the local value of a trace is structure-shared with the slash value, constraints on the trace will be constraints on the filler. For each nonlocal feature, the inherited value on the mother is the union of the inherited values on the daughter minus the to-bind value on the head daughter. • For example, hates specifies that its object be accusative, and this case information is local • In other words, the slash information (which is part of inherited ) � synsem | local | cat | head | case acc � • So, the trace has as part of its percolates “up” the tree entry, and thus the filler will also have to be accusative • This allows the all the local information of a trace to “move up” to the (21) *He i /Him i , John likes i filler 13 14 The top of a UDC: Filler-head structures The top of a UDC: Filler-head structures Filler-head schema Explanation of the schema � phrase � → • Filler and trace are identified as the exact same thing (as far as their dtrs head-filler-struc local structure is concerned)       � verb � • The trace is “bound” by the to-bind feature; this prevents the slash  head  loc | cat vform fin  value from going any higher in the tree             subcat ��  head-dtr | synsem   • Only saturated finite verbs (i.e., sentences) license such structures       � �  � �  inherited | slash element  1    nonloc     � � to-bind | slash 1     filler-dtr | synsem | local 1 15 16

The analysis of the strong UDC example S ˆ nloc | inherited | slash {} ˜ f h The top of a UDC: Filler-head structures NP S Example for a structure licensed by the filler-head schema " " # # inherited | slash ˘ ¯ ˆ ˜ 3 local 3 nloc ˘ ¯ to-bind | slash 3 c h John i NP VP � � " " # # nloc | inherited | slash {} inherited | slash ˘ ¯ 3 nloc to-bind | slash {} we h c f h V S � � � � � � � � 2 loc | cat | subcat �� 3 inherited | slash . . . , 1 ,. . . local 1 " # nloc inherited | slash ˘ ¯ know 6 3 7 � � 4 nloc 5 to-bind | slash 1 to-bind | slash {} c h 1 NP VP 2 loc | cat | subcat ˙ ¸ 3 1 " # inherited | slash ˘ ¯ she 3 6 7 4 nloc 5 to-bind | slash {} h c V NP " # " # loc | cat | subcat ˙ ¸ 1 , 2 loc 3 17 2 nloc | inher | slash ˘ ¯ nonloc | to-bind | slash {} 3 likes i The analysis of weak UDCs Lexical entry of adjective easy (22) a. Kim i is easy (for John) to please i 2 phon < easy > 3 b. Kim i is easy to prove that Mary asked Paul to bribe i . 2 3 2 2 3 3 adj 6 head 7 6 7 “ ” 6 7 (23) a. It is easy to please him acc / * he nom . 6 6 7 7 6 * NP 1 , ˆ for ˜ + 7 PP 3 , 6 7 6 cat 7 6 6 7 7 6 6 7 7 6 subcat 7 6 7 6 7 6 h “ ” i 7 b. I nom am easy to please acc . 6 4 5 7 ˆ acc ˜ 6 VP inf , inher | slash element 2 NP : ppro 1 : 4 7 6 7 6 7 6 6 7 7 6 7 loc 6 7 6 7 6 7 2 3 6 synsem easy 7 6 7 6 7 6 6 7 7 6 7 6 6 7 7 arg1 1 ref ⇒ No true (non-argument) filler, only coindexed items serving as arguments 6 6 7 7 6 7 6 7 6 cont 7 6 7 6 6 7 7 6 7 6 arg2 3 7 6 6 7 7 4 5 4 5 6 7 6 7 arg3 4 6 6 7 7 4 5 4 5 Subject is role assigned: ˘ ¯ nonloc | to-bind | slash 2 (24) a. I believe there to be a unicorn in the garden. ⇒ Lexical entry selects for infinitive complement missing an NP, which is b. * There is easy to believe a unicorn in the garden. coindexed with the subject (25) a. [ This sonata ] i is easy to play i on that violin. b. [ This violin ] i is easy to play this sonata [ on i ] . 19 20